Subband adaptive equalization of time-varying channels

The potential presence of fractional delays, nonminimum phase parts, and a colouring of the channel output can require very large adaptive equalizers exhibiting slow convergence for LMS-type algorithms. In this paper, we address the equalization problem via a subband approach, which can significantly reduce computational complexity and improve convergence speed. Of particular interest in this paper is the tracking behaviour, which determines the speed with which the algorithm is capable of following time variations in the channel. We give reasons why tracking of LMS-type subband adaptive algorithms should be comparable to the full-band case. Simulation results with a time varying channel are presented.